Tarsney, Christian (2019). Rejecting Supererogationism. Pacific Philosophical Quarterly 100 (2):599-623.
Even if I think it very likely that some morally good act is supererogatory rather than obligatory, I may nonetheless be rationally required to perform that act. This claim follows from an apparently straightforward dominance argument, which parallels Jacob Ross's argument for 'rejecting' moral nihilism. These arguments face analogous pairs of objections that illustrate general challenges for dominance reasoning under normative uncertainty, but (I argue) these objections can be largely overcome. This has practical consequences for the ethics of philanthropy -- in particular, it means that donors are often rationally required to maximize the positive impact of their donations.
I have not fully evaluated the arguments in this paper, but I would be pretty interested and surprised to learn that we cannot justify the existence of supererogatory actions. I lean towards virtue ethics, and there supererogatory seem acceptable: you might be able to express greater virtue by taking actions that produce more good, but intuitively it seems there should be some "baseline" of virtue which, if you meet it, is good enough. That is, it seems reasonable to me to permit people to be not maximally virtuous as long as they are sufficiently virtuous as to not cause harm. But Tarsney argues, at a very high level gloss, that moral uncertainty makes superergatorism untenable and we must always perform the best action or fail to maximize expected value. Perhaps we disagree primarily on the aim to maximize expected value?
From the conclusion:
I have argued that, when an agent is uncertain whether some practical option is obligatory or all-things-considered supererogatory, even if she thinks it is most likely supererogatory, she may nevertheless be rationally required to choose that option. And I have argued that, according to the most plausible forms of supererogationism, ordinary cases of moral supererogation are also cases of all-things-considered supererogation. The major qualification that must be added to these conclusions, as we have seen, is that the agent must not have credence greater than .5 in satisficing versions of supererogationism that endorse strongly permissive principles like EPP – that is, unless it turns out that the normative force of dominance principles like GDoT* or FDoT is simply independent of the agent’s beliefs. But because principles like EPP are implausible – and less plausible, I think, than the versions of nihilism that threaten to undermine Ross’s argument – it seems to me that the dominance argument for rejecting supererogationism is on a surer footing than the dominance argument for rejecting nihilism.
The practical scope of the anti-supererogationist argument depends mainly on how often we can be certain that some potentially-supererogatory option is at least as objectively choiceworthy as its less morally attractive alternative, e.g., that donating to charity is at least as choiceworthy as spending the same money on luxury goods for myself. If one is disposed to think that we should assign positive credence, however slight, to every or nearly every proposition, then of course dominance principles will have no scope at all for application. But this limitation of dominance reasoning is no more operative in the context of moral uncertainty than in the context of empirical uncertainty – if anything, less so, since it seems more plausible that we should always have positive credence in all empirical propositions than that we should always have positive credence in all moral propositions. Dominance principles in their native, game-theoretic context depend for their relevance on an idealized assumption of partial certainty that rarely if ever describes our real choices.